[ENH] Implement GaussianRegressor using scikit-learn's LinearRegression adapter

[sanjayk0508]

towards #7

This PR introduces Gaussian regression implementation using scikit-learn's LinearRegression, fulfilling the basic capabilities of Gaussian regression within the skpro.

Key Changes

• Implemented the GaussianRegressor class, which adapts scikit-learn's LinearRegression for Gaussian regression.
• Added methods for model fitting, prediction, model evaluation, parameter handling, and testing support.
• Aligns with existing skpro architecture and conventions.
• Comprehensive documentation

Additional Notes

• Tested with pytest for the implementation to ensure correctness and reliability.

I think there's still potential for more enhancements, this PR sets the groundwork for GaussianRegressor. Open to feedback and suggestions for further refinement and iteration.

[fkiraly]

Thanks.

This could in-principle be a useful addition, but it does not comply with the extension template, see regression.py.

In particular, you should:

• not override predict and predict_proba. Implement _predict and _predict_proba.
• _predict_proba should return a probability distribution object, an skpro BaseDistribution.
• predict should not write any attributes to self.
• I think you can't use _DelegateWithFittedParamForwarding, because LinearRegression.predict does not have a return_std argument

I would recommend to start with the extension template, not with the sklearn adapter template. That is, extension_templates/regression.py.

How about you look at statmodels GLM instead, with a Gaussian link?
https://www.statsmodels.org/stable/glm.html#module-statsmodels.genmod.generalized_linear_model